0 évaluation0% ont trouvé ce document utile (0 vote)
22 vues11 pages
Abstract In this paper, we make a preliminary attempt to approach the phenomenon of prospective memory (PM) from the point of view of Haken’s theory of synergetics that refers to complex, self-organizing systems, in gen- eral, and to brain functioning and cognition, in partic- ular. In the following, we consider one form of PM only—the so-called event- or cue-dependent PM. We first interpret cue-dependent PM in terms of synergetics and then applythe mathematical formalism of syner- getics.
Titre original
A synergetic interpretation of cue-dependent prospective memory
Abstract In this paper, we make a preliminary attempt to approach the phenomenon of prospective memory (PM) from the point of view of Haken’s theory of synergetics that refers to complex, self-organizing systems, in gen- eral, and to brain functioning and cognition, in partic- ular. In the following, we consider one form of PM only—the so-called event- or cue-dependent PM. We first interpret cue-dependent PM in terms of synergetics and then applythe mathematical formalism of syner- getics.
Abstract In this paper, we make a preliminary attempt to approach the phenomenon of prospective memory (PM) from the point of view of Haken’s theory of synergetics that refers to complex, self-organizing systems, in gen- eral, and to brain functioning and cognition, in partic- ular. In the following, we consider one form of PM only—the so-called event- or cue-dependent PM. We first interpret cue-dependent PM in terms of synergetics and then applythe mathematical formalism of syner- getics.
A synergetic interpretation of cue-dependent prospective memory Received: 14 September 2004 / Revised: 6 December 2004 / Accepted: 9 December 2004 / Published online: 23 February 2005 Marta Olivetti Belardinelli and Springer-Verlag 2005 Abstract In this paper, we make a preliminary attempt to approach the phenomenon of prospective memory (PM) from the point of view of Hakens theory of synergetics that refers to complex, self-organizing systems, in gen- eral, and to brain functioning and cognition, in partic- ular. In the following, we consider one form of PM onlythe so-called event- or cue-dependent PM. We rst interpret cue-dependent PM in terms of synergetics and then applythe mathematical formalism of syner- getics. Introduction Generally speaking, prospective memory (PM) refers to a remembered future. This is in contrast to retrospective memory that refers to a remembered past (Neisser 1982). More specically, PM studies commonly focus their attention on the realization of delayed intentions or plans, so that the task of PM involves the future use of memoryone remembers to remember and then per- forms accordingly (Sellen et al. 1997). Examples of PM are abundant as the phenomenon is central to the daily experience of every individual: the plan to shop in a supermarket after work, the intention to send a letter when seeing a mailbox, to buy owers on ones way to a friend, to call a colleague at 3 p.m. and so on. Examining the various cases, students of PM have made a distinction between time-dependent and cue- dependent PM (Brandimonte et al. 1996; Marsh and Hicks 1998). An example of the rst is the intention to call a friend at 3 p.m. as promised, or to take a medicine every 4 h; an example of the latter, to buy owers when passing a ower shop. The case of the plan to shop in the supermarket after work is related to both time and cue/ event: returning home after work might be treated as an event or cue, but it is also related to time. An important aspect of PM is that during the time lag between the moment of intention and the moment of performance of the PM task, the person temporarily forgets about (that is, is not aware of) the task and can engage his or her memory in many other important operations (Smith 2003). Then, at the proper time, place, or event, one suddenly remembers: to shop, buy, call a friend or take a medicine. One can say that in this sudden event the remembered PM task is retrieved from long-term to working memory store. Of specic rele- vance here is the intention superiority eect (Goschke and Kuhl 1993) that refers to the nding that retrieval rates of the stored intentions are better than average (Marsh et al.2002a, b). As in other memory tasks, so in PM, one often remembers the PM task and then performs it at the right time or place, but quite often one might also forget about it altogether or remember to perform the task too late or too early with the consequence that the realiza- tion of the planned action is not optimal (Graf and Uttl 2001). Apparently, this might be a consequence of the fact that the transfer of a delayed intention/plan from long-term to working memory involves complex rela- tions between the planned PM task and the activities executed during the delay period. These complex rela- tions might take the form of competition when one memory task suppresses the other and they might take the form of cooperation when one memory activity reinforces the other. For example, the aim of a research H. Haken (&) Institute for Theoretical Physics, Center of Synergetics, Pfaenwaldring 57/4, 70550 Stuttgart, Germany E-mail: haken@theo1.physik.uni-stuttgart.de Tel.: +49-711-6854990 Fax: +49-711-685-4909 J. Portugali ESLabEnvironmental Simulation Laboratory (Porter School for Environmental Studies), Department of Geography and the Human Environment, Tel Aviv University, Tel Aviv, 69978, Israel E-mail: juval@post.tau.ac.il Tel.: +972-3-6408661 Fax: +972-3-6406243 Cogn Process (2005) 6: 8797 DOI 10.1007/s10339-004-0041-z currently under way (Mintz et al., unpublished) is to test whether the navigation in 3D virtual environments fea- turing dierent visual load suppresses or reinforces time- dependent PM tasks. Despite the centrality of PM in everyday life, the number of studies on this phenomenon is relatively small and until recently the progress made in the eld has been relatively slow (Sellen et al. 1997). Recently, however, the interest in a systematic study of PM seems to be growing (Bradimonte et al.1996; Ellis and Kvavilashvili 2000). Many PM studies focus their attention on the decline of PM capability with age and disease (Reese and Cherry 2002; Einstein and McDaniel 1990; Einstein et al. 1998; Ellis and Kvavilashvili 2000; Smith et al. 2000; Bravin et al. 2000; West et al. 2002; Vogels et al. 2002), a few on the cognitive structure of the PM process (Marsh et al. 2002a, b; Ellis and Kvavilashvili 2000) and on the brain structures involved (Burgess et al. 2003, 2001; Ellis and Kvavilashvili 2000; West et al. 2001). We were not able to nd formal theorizations of PM. In this paper, we make a preliminary attempt in this direction, namely we approach PM from the point of view of Hakens (1983, 1987) theory of synergeticsin particular from its pattern recognition paradigm. In the following, we will rst interpret PM in terms, of synergetics and then suggest a formalism to one form of PM onlyto cue- dependent PM. Before we do so, however, we need to say a few introductory words about synergetics. A short introduction to synergetics Synergetics is the name of Hakens (ibid) theory of self- organization. The notion self-organization itself is a fundamental property of open, complex and far from equilibrium systems. Open is used in the sense that they exchange matter, energy and information with their environment. Complex is used in two respects: in the sense that their nonlinear components are intercon- nected by a complex network of feedback and feedfor- ward loops, and, in that the number of their parts is large to the extent that there is no way to dene causal relations among them. Far from equilibrium is used in the sense that such systems are never at rest, never in equilibrium, but rather in a continuous movement typ- ied by oscillations, rhythms, chaos, uctuations as well as steady states, depending on specic conditions (control parameters). Such systems can exhibit phe- nomena of circular causality, nonlinearity, instability, spontaneous formation of spatial, temporal, functional structures including fractals and chaos. In particular, they attain their order spontaneously, i.e. by means of self-organization. (For a concise introduction and bib- liography, see Portugali 1999, Chaps. 3.1 and 13.1). As the name indicates, the synergetic approach to complex systems puts special emphasis on the interac- tion between the many parts of the system. According to synergetics, this interaction, which is often regulated by the so-called control parameter, gives rise to several or- der states that enter into competition. The latter is re- solved when one (or a few) of the order states wins the competition and enslaves the many parts and compo- nents of the system to its specic movement. This win- ning state is termed order parameter and the above process of enslavement , the slaving principle. The general scenario suggested by synergetics is thus a pro- cess of circular causality by which local movements and interactions give rise to a global systemic order and movement that in turn determine local interactions and so on. The theory of synergetics was originally formulated in physics in the domains of laser and uid dynamics and with two types of formalisms known as the laser paradigm and the uid dynamic paradigm. In the last three decades, however, the theory was applied to sev- eral other domains such as sociology, economics, urbanism, and most intensively to the study of brain functioning, cognition and behavior (Haken 1990, 1991, 1996; Kelso 1995). The latter application entailed two more formalisms known as the pattern recognition par- adigm, and the nger-movement paradigm. The para- digm that is specically relevant to our present discussion on cue-dependent PM is the pattern recog- nition paradigm. A typical pattern recognition experiment starts when a test person (or a computer) is oered a small portion from a whole pattern and is asked to recognize it out of a repertoire of patterns that are stored in memory (Haken 1983). The classical example here was face recognition (Fig. 1) but other types of patternsmaps, hand-written letters, etc.were also experimented on. The synergetic theory and algorithm of pattern recognition were developed by an analogy to its former theory of pattern formation. The logic behind this analogy is that pattern recognition is pattern formation in the mind. According to synergetics, the few parts of a pattern oered to the test person enter into interaction, out of which, by means of associative memory, an order parameter Fig. 1 Pattern recognition of a simple scene (one faces and a background). Top Examples of stored prototype including the encoded family names. Bottom Examples of a recognition process where part of a face is prescribed as initial state 88 emerges. The order parameter enslaves the rest of the parts and recognition is established. As can be seen, the notion of associative memory stands at the core of the pattern recognition process. Intuitively speaking, this notion refers to the tendency that when one sees a pattern, say a persons face, one wants to know the name of that person or the past cir- cumstances under which one rst met him/her. In other words, one wishes to associate one dimension of the patternthe facewith its other dimensions: name, event of meeting, place of work and the like. Associative memory thus refers to the process by which an incom- plete set of data yields the full set (Haken 1996, p 239). In the various pattern recognition experiments, one can make a distinction between simple versus complex scenes. Figure 1 is an example of a simple scene: it is composed of a single semantic entity or a category (a face) and a background. A more complex scene is shown in Fig. 2. Here, we have two faces and a background. The theory of synergetics suggests that the recognition of such scenes evolves by means of a sequential play between the so-called attention parameters. The mind/ computer rst directs attention (parameter) to the rst face, recognizes it, then to the second and so on. As will see, attention parameters are specically important in the execution of cue-dependent PM tasks. Another relevant distinction is between pattern recognition of static versus dynamic patters, that is to say, patterns of action. An example of the rst is the pattern recognition of the faces discussed above (Figs. 1, 2), while of the second a moving person or animal (Fig. 3). Patterns of action have been studied in some detail in the context of synergetics. These studies have shown, rstly, that patterns of action evolve as a self-organizing system by which the coor- dination between the large number of elements in- volved in the process (ranging from neurons to muscles and joints, etc.) is governed by one or a few order parameters. Secondly, that likewise dierent patterns of action/movement (walking, running, etc.) are governed by dierent order parameters. Thirdly, as a consequence, the transition between one pattern of action to the other involves a phase transition between one order parameter to the other. Fourthly, the pro- cess of learning a certain pattern of action is a process by which the large number of elements involved in the activity are being enslaved and coordinated by one or a few order parameters. This latter process of learning was studied in some detail with respect to the use of a pedalo (Fig. 4). The above process of pattern recognition can be and has been implemented on the synergetic computer, which is a synergetic network model of pattern rec- ognition (Haken 1991). The synergetic computer can be realized by a three-layer network as described in Fig. 3 Two patterns of action/movement dominated by two dierent order parameters. Top Gallop. Bottom Canter Fig. 4 Schematic drawing of a pedalo. The pedalo consists of two steps that are eccentrically mounted on a wheel. The resulting movement of the left and right leg is shifted by 180. The right-hand side shows the side Fig. 5 Diagrammatic illustration of the synergetic computer Fig. 2 Pattern recognition of a complex scene (two faces and a background) 89 Fig. 5. The uppermost layer contains the individual neurons with their activities that by means of their interaction give rise to the order parameters and their dynamics that are described by the middle layer. The one or few order parameters that eventually enslave the many parts of the system govern the activity pattern of the neurons represented by the lower layer. In cases of patterns of action, we may identify the uppermost layer with sensory neurons and the lower layer with motor neurons. This synergetic computer will be employed also in our interpretation of PM in the remaining part of this paper. From pattern recognition to PM Applying the pattern recognition paradigm to the case of prospective memory involves three new adaptive features: the rst proposes a link between retrospective and prospective memory. The second suggests an analogy between the cue of PM and the small part of a whole pattern oered to a test person in the pattern recognition process. The third suggests an interpreta- tion of the cue-dependent PM in terms of a compe- tition between existing and prospective attention parameters: both are competing for control over working memory. The link between retrospective and prospective memory In pattern recognition, the test person (or computer) recognizes the complete meaning of a partial portion of a pattern oered to him/her/it by means of a rep- ertoire of patterns stored, or rather constructed, in retrospective memory. Such learned patterns, con- structed as they are on the basis of the persons past experience, might refer to a specic face the test per- son encountered in the past, to more general patterns such as a human face, a building, a car, as well as to patterns of action and behavior such as crossing a busy road, shopping in a supermarket, commuting to work and back home and so on. In the synergetic computer, each such pattern is described by means of an order parameter. We suggest that the prospective memory process starts when the person employs his/her past experience and previously constructed similar patterns of action in order to construct in mind a pattern of action that refers to an imagined or virtual future. Such intended, plan- ned, or imagined pattern of action might be making a phone call, shopping in a supermarket, or a scenario composed of a sequence of actions such as start com- muting home, get o the road at a certain junction, park the car, shop in the supermarket etc. etc. As in ordinary pattern recognition, here too, in the synergetic computer each such virtual pattern is described by means of an order parameter. The cue as a partial pattern An ordinary process of pattern recognition starts, as noted above, when the test person is oered part of a pattern and is asked to recognize it out of a repertoire of whole patterns stored in memory. The pattern recogni- tion process is thus being triggered by the partial pattern oered. In light of what has been said above, the object, or event, or spatial scene that triggers a cue-dependent PM can be treated as part of a planned pattern of action that makes the PM task. For example, the mailbox that one suddenly encounters when driving from home to work is part of a pattern of action that involves, say, parking the car, crossing the road, putting the letter in the box, crossing back the road, getting into the car and driving on to work. As in pattern recognition so in PM, a small part or segment from the larger pattern entails a recognition process that in the case of PM includes also the execution of the planned action. The role of attention parameters In ordinary pattern recognition experiments, the test person is asked to recognize an oered pattern. The person is thus forced to direct attention to the pattern and the recognition task. In cue-dependent PM experi- ments, the subject is shown a partial oered pattern but is not asked, (nor demanded or reminded) to respond to it. In fact, the question in PM experiments is whether and in what circumstances the subject will respond to the cue or will forget about it. From the point of view of synergetics, the shown pattern that functions as a cue is a sort of attention parameter that must compete with the attention parameter that preoccupies the persons attention at the timespace moment of the cue. We suggest that the above noted experiments with pattern recognition of complex scenes give us insight into how the PM task is implemented. As we have seen, the recognition process of complex scenes is sequential: the attention parameter rst draws the Fig. 6 A complex scene with a fair competition between two order parameters corresponding to two patterns, faces versus vase 90 mind-brain/computer to one element of the scene, recognizes it, then the attention parameter is put on the second element and so on. A somewhat dierent situation occurs with respect to the ambiguous vase faces patterns (Fig. 6). Here, there are two attention parameters oscillating between attention to vase versus attention to faces. The latter case can be interpreted as a competition between two order/attention parameters: vase versus faces. The competition between the two attention parameters might be fair, but in many cases it is from the start biased as in the ambivalent pattern young/old lady of Fig. 7. Here, 80% of the male subjects rst recognized the young lady and only later the old one (Haken 1996). The above biased situation is specically relevant in PM tasks. Here, too, there is competition between two order/attention parameters: the PM attention parameter versus the one that dominated the subjects mind at the spacetime moment of the cue. One can say that the process starts when the subjects attention is occupied/ biased by a certain task and the role of the cue is to create a stronger biasone that will direct the persons attention to the PM task. If the bias/eect of the cue is weak, the person will forget; if it is strong, it will enslave the subjects mind and the task will be executed. As noted above, Goschka and Kuhls (1993) intention superiority eect indicates that retrieval rates of the stored intentions are better than the average (Marsh et al. 2002a, b). A synergetic model of cue-dependent prospective memory Let us rst reconsider a typical cue-dependent PM task: we may plan to buy some medicine but do not have a specic plan how to do that. Then, while driving through the city, suddenly we see the word pharmacy whereupon our prospective memory becomes active. We stop our car, park it, go to the pharmacy and buy the medicine. This example provides us with the main fea- tures of our approach. First, we must prepare our ac- tion, then we must be able to recognize a specic cue and nally we will carry out our actions. More specically, we suggest interpreting the PM task in terms of three phases, namely the preparatory phase, trigger phase and execution phase. In what follows, we will analyze them in more detail. We will rst provide the reader with a qualitative description of the dierent phases and then give an explicit mathematical description. To make the article understandable to a wider audience, we make this mathematical description as explicit as possible, whereby we focus our attention on simple models. It should be stressed, however, that based on such an approach the mathematical treatment can be pushed much further. Indications of such developments will be given below at the proper places. The mathematical tools and concepts will be taken from synergetics. Qualitative approach Preparatory phase (A1) In the process of cue-dependent prospective memory, we are dealing with a twofold learning process that comprises, rst, retrospective action patterns learned in the past a j , j=1,...,n (by actual doing) and governed by order parameters n j (t). Second, and spe- cically signicant to the present issue, prospective mental preparation of virtually learned patterns v k , k=n+1,...,N governed by order parameters n k (t), k=n+1,...,N (imagination of action patterns that were not yet performed). As many of the cues in cue-dependent PM tasks are related to specic environments and objects in ita specic junction or a shopping center in a certain city, for examplethe above retrospective learned patterns might include the cognitive map of that city or relevant parts of it (Portugali 1996). As we shall see, the whole process will be strongly inuenced by attention param- eters so that we must consider the following: (A2) setting of thresholds of attention parameters k l thr, k=1,...N We must be aware of the fact that attention param- eters depend on our general disposition including our hormone level, fatigue, etc. Trigger phase (A3) During this phase, we perform a conscious or subconscious search for patterns p m , m=1,...,M that may serve as cues p c l n for the attention parameter k l of the desired action a l . Note that there may be several cues labelled by l for the same attention parameter related to the desired action a l . Having in mind the example given at the outset of this section, we may look for the word pharmacy and the desired action will be buying a specic medicine. As we shall see below, this search is connected with the last point of Sect. A1. Fig. 7 A complex scene with a biased competition between two order parameters corresponding to two patterns, young woman versus old woman 91 Execution phase The execution phase has two parts: in the rst part, we perform actions that are not included in the cue- dependent prospective memory, e.g. we are driving by a car in the city or we are walking along a street. During this period, one or several order parameters that govern our ongoing present action are active. But then at a certain instant, the cue-dependent prospective memory becomes active and the corresponding specic action starts. For this to happen, we require: (A4) that the attention parameter for this specic new action is larger than its threshold, i.e. k l 0 > k thr l 0 ; all other k l \k thr l In this case, a specic action control parameter g l increases from zero and governs via an order parameter n l the action a l by enslaving the subsystems necessary for that action. During a transition period, the other order parameters that have been active so far decline and are no more active. Note that if the above attention parameter for the PM cue is not larger than the threshold, the PM task will not be performed. In other words, the persons attention will continue to be enslaved by the pattern of action driving or walking in the street and the person will forget about his/her PM pharmacy task. Let us now turn to the mathematical formulation. Mathematical formulation In order to facilitate the reading of the mathematical part, we refer the reader to Diagram 1, in which the 92 roman numbers indicate the individual steps, and to Fig. 8 that illustrates each step by means of a picto- gram. Step I (B1) The actions a j , v k may be steered by motor pro- grams that are executed by a computer. According to, the traditional concept of brain functioning, our brain steers movements by means of a motor program in analogy to a computer. Over the past years, however, a paradigm shift, triggered by the Haken et al. (1985) model, has taken place in that the steering of movements by the behavior brain is conceived as a process of self- organization governed by order parameters. Since in this paper we are primarily interested in the behavior of humans, we will stick to the latter approach in the fol- lowing. These self-organized actions may be quite complex and may be composed of a series of elementary actions. For clarity, we restrict ourselves to very simple examples where we deal with one kind of action gov- erned by one or several order parameters. More com- plicated actions will be visualized, at least in our approach, by a series of individual elementary actions where the switch from one action to the next one is governed by so-called control parameters. One of sim- plest and yet rather realistic order parameter equations has the form: dn dt an bn 2 cn 3 This is a dierential equation for the real order parameter n with constant coecients a, b, c. In the case of a periodic movement such as walking or running or driving a pedalo, the order parameter n becomes com- plex, i.e. Fig. 8 Pictograms corresponding to steps IVIII in diagram 1 93 n n r in i where i is the imaginary unit and n r and n i are the real and imaginary parts of n, respectively. In such a case, the typical order parameter equation reads dn dt an bn 2 cn n j j 2 where the coecients a, b, c may now be complex and in particular a c ix where c is a constant and x the frequency of the overall motion. Equations of this type were studied in detail for driving a pedalo (see Fig. 4, above) and for walking and running (see above). The actually observed movement is described by angles J j (t) between the limbs, such as upper and lower leg, upper and lower arm, and the body. The temporal change of all these angles is gov- erned by a single order parameter n or, in mathematical parlance, the angle J j (t) is determined as # j t f j nt where f j is a specic function. We may even state that speaking can be modelled the same way. In order to give the reader an impression of the wide range of applicability of the order parameter concept, we just mention that order parameter equa- tions were explicitly established and studied in detail in the case of quadruped gaits, such as walk, trot, and gallop of horses and other kinds of gaits, say of camels etc (Haken 1996, Chaps. 912). In particular, this approach allows one to study the transition be- tween various gaits. In general, the order parameter equations are of the form: dn l dt N l n l ; g l F n l l t 1 n l , N l may be vectors and g l is a control parameter that, as usual in synergetics, indirectly controls the movement based on self-organization. F n l l t is a uctuating force that is needed to initiate the action l provided that the control parameter possesses a value larger than a critical one. In many cases, but not exclusively, the always non- linear functions N l may be written as a low order poly- nomial with constant coecients. Numerical values of these coecients determine the kind of time dependence of the order parameters, which jointly with the slaving principal xes movements of, e.g., limbs. At a still more microscopic level, these order parameters determine also the action of the individual neurons, e.g., their coordi- nated ring rates. Step II (B2) We now turn to the xation of the thresholds k l thr for the attention parameters k l that will appear below as a specic kind of control parameters in the action con- trol parameter equations. We want to secure that the prospective memory, with the action then caused by it, will become active only if the cue is strong enough. Therefore, we shall assume k thr l 0 C > 0 2 and all other k thr l C 0 with C 0 > C l 6 l 0 : 3 Steps IIIV (B3) In the next steps, we shall deal with the dynamics of the attention parameters k l . We rst assume that the attention parameters are such that the original action is performed, e.g., walking along a street (step III). But then a new object related to the PM is observed (step IV). In such a case, a competition, between attention parameters sets in (step V). To describe this competition, we treat the attention parameters formally as order parameters as in the case of pattern recognition. The corresponding equations read: dk l dt K l Ak 2 l B X j k 2 j ! k l F k;l t: 4 A and B are constants. F k l are again uctuating forces. Equation 4 takes care of our idea that atten- tion works like a kind of search light and may jump from one object (outside or inside our brain) to an- other one. The competition is mainly governed by two eects, namely: 1. The dependence of L l on l whereby L l may be time dependent. It may depend on interesting patterns to be recognized, which leads to a sensitization of overall pattern recognition, but also saturation (fati- gue) may be taken into account by letting L l decay for all l. 2. The second ingredient that especially has to take care of the eect of the cue-dependent prospective mem- ory must be a fast dynamics by which the preselected l is activated and thus L l becomes large enough. Equation 4 must be complemented by initial condi- tions. In all cases, we assume that except for the new action to be undertaken all the other attention param- eters have a specic value at the time of the recognition of the cue. In this case, we choose: K l 0 c X l;k 00 p l tp c k 00
2 ; k 00 2 l 0 ; 5 where c is a proportionality factor. The motivation for this equation is as follows: while driving in the city or walking in the street, we observe patterns of all sorts all the time, patterns that 94 we distinguish by an index l. As usual in pattern recognition, the pattern is encoded by means of a pattern vector p l . However, because of our movement including the movement of our head, this pattern vector that we observe is time-dependent. It appears for instance, becomes more or less complete and vanishes again. These patterns are to be compared with the learned or virtually learned cue pattern that we denote by p k c . These are the patterns that serve in the sense of pattern recognition as prototype patterns. On the other hand, they are those patterns that will lead to the action a l . Jointly, with Eq. 4, the result will be as follows. The brackets in Eq. 5 denote the scalar product between the vectors. There are two more alternatives to Eq. 5, namely we further drop the sum over l and just use at time t the actually seen pattern p l . Our nal alternative and prob- ably the most convenient one is: K l 0 c X k 00 ptp c k 00
2 k 00 2 l 0 pt 2 p l 6 where p(t) is the pattern actually seen at time t. The relation 6 means:L l >0 if p(t) contains one of the cue patterns and =0 otherwise. In order that L l wins the competition, L l must be larger than all other L l . Steps VIVIII (B4) We now turn from attention to the transformation into action. To this end ,we use the action control parameters (as special case: movement control parame- ters) g l . We now admit that they are time dependent and obey the competition equations dg l dt k l k thr l g l ag 2 l b X K g 2 K ! g l F g l l t 7 a and b are constants and F g l l uctuating forces. For what follows, we distinguish between slow and fast dynamics of g l . Here, slow and fast refer to the tem- poral changes of g l (t) as compared to that of the movement of the enslaved parts (e.g., the limbs). In the slow case, it will be sucient to use Eq. (1) that must be supplemented by the slaving principle that now in its leading approximation reads # j t X l g l tf j;l n l t: 8 It means that the movement coordination of the individual parts is a sum over the individual contribu- tions f j,l with weights g l . In higher approximations, we must invoke a general relationship # j t G j g 1 t; :::g n t; n 1 t; :::n m t 9 This form takes care of the fact that, for instance, 0on nerve signals etc. For our present approach, Eq. 8 seems sucient, however. The meaning of Eq. 7 jointly with Eqs. 1 and 8 or Eq. 9 is as follows: Before the cue had been recognized, a series of actions may have occurred (e.g., driving the car in a specic street; cf. step III). Let us call the last action l 0 . Then, g l 0 was the winner of Eq. 7. Now k l is switched to its high value (step V), and in slow transient g l 0 vanishes, whereas g l grows to its maximal value (step VI). During this period, the dynamics described by n l 0 t is switched o (all others were zero or small anyway) and that of l is switched on according to Eq. 1 (step VII). Concomitantly, the movement of the enslaved parts (e.g., limbs) switches smoothly from mode l 0 to l (step VIII). Remarkably, during all this process, the actions can be exclusively described by the order parameters n l . In the case of fast dynamics of g l , the situation (probably) changes. Here, specic initial conditions of the enslaved parts must be taken into account, which requires a detailed study of the movements of the indi- vidual parts based on more degrees of freedom than those provided by the order parameters. This situation is beyond the scope of our paper, because it would become too specialized otherwise. In specic situations that require a fast reaction in typied situations (e.g., in playing tennis), the execution of a movement must very quickly follow the observation of a specic cue. In such a case, we may encounter a shortcut in the sequence of steps IVVII by directly putting g l k l and dropping Eq. 7. Generally speaking, the sequence of steps IVVII may be denoted as sensory- motor action. Instantiating the model Given the above model, one can now explore specic event based PM phenomena. For example, empirical studies on the cue-dependent PM have found that dif- ferent types of cues have dierent eects on the perfor- mance of cue-dependent PM (Marsh et al. 2000; Taylor et al. 2004; Van Der Berg et al. 2004 and references therein). Salient cues have been found to be more eective than non-salient ones, specic cues entailed better performance than categorical cues, environmen- tally prominent cues elicit better performance than less prominent ones and so on. The model suggested above allows one to study such phenomena. In order to do so, however, one should rst nd a way to quantify salient and non-salient cues, strong versus weak cues and so on. We suggest doing so by making a link to our previous study the face of the city is its information (Haken and Portugali 2003). Commencing from a conjunction between Shannons information theory (Shannon and Weaver 1949) and Hakens (1988/2000) self-organization approach to information, this study shows, rst, that dierent ele- ments in the environment aord dierent levels of information. Second, that the amount of information aorded by the various elements can be quantied by means of Shannon information bits. In light of our 95 information study, we put forward the hypothesis that dierent types of cues aord dierent quantities of information bits that in their turn entail corresponding eects on performance in cue-dependent PM. The next step is to introduce the cues aording dif- ferent quantities of information bits to our prospective memory model. This will immediately aect the role of attention parameters in the various phases. First, at the A2 preparatory phase, when we set the thresholds of attention parameters k l thr, k=1,...N . Next, in the A3 trigger phase, when we perform a search for patterns p m , m=1,...,M, that may serve as cues p c l n for the attention parameter k l of the desired action a l . Finally, at the A4 execution phase, where we make sure that the attention parameter for the specic PM action is larger than the thresholds set at the A2 phase, that is, k l >k l thr , all other k l <k l thr . Concluding remarks In our paper, we have tried to cast the phenomenon of cue-dependent prospective memory into a conceptual form, which can furthermore be cast into a mathemati- cal description that eventually can be implemented on the computer. More specically, the above formalism was developed in a way that allows a realization of the parallel synergetic neural network specied by Haken (1991) in Synergetic Computers and Cognition, as well as on a serial computer (e.g., a PC) whose output mimics the process of self-organization. In this way, it might serve as a model of a real brain that operates by means of self-organization (e.g., eects of control parameters) rather than by computer programs. In the future, we intend, rst, to relate our model to empirical studies on the cognitive structure of the PM process that seem to comply with our approach (e.g., Marsh et al. 2000, 2002a, b; McDaniel and Einstein 2000). Second, to extend the model in a way that will take into account the spatial dimension of cue-dependent PM (e.g. Mintz et al., unpublished)an issue that is central to the PM phenomenon but has not as yet received sucient treatment. Third, we in- tend to develop a synergetic model for time-dependent PM. References Brandimonte M, Einstein GO, McDaniel MA (eds) (1996) Pro- spective memory: theory and applications. Lawrence Erlbaum, Mahwah Bravin JH, Kinsella GJ, Ong B, Vowels L (2000) A study of per- formance of delayed intentions in multiple sclerosis. J Clin Exp Neuropsychol 22(3):418429 Burgess PW, Quayle A, Frith CD (2001) Brain regions involved in prospective memory as determined by positron emission tomography. Neuropsychology 39(6):545555 Burgess PW, Scott SK, Frith CD (2003) The role of the rostal frontal cortex (area 10) in prospective memory: a lateral versus medial dissociation. Neuropsychology 41(8):906918 Einstein GO, McDaniel MA (1990) Normal aging and prospective memory. J Exp Psychol Learning Memory Cogn 16(4):717726 Einstein GO, McDaniel MA, Smith RE, Shaw P (1998) Habitual prospective memory and aging: remembering intentions and forgetting actions. Psychol Sci 9:4 Ellis J, Kvavilshvili L (2000) Prospective memory in 2000: past, present, and future directions. Appl Cogn Psychol 14:19 Goschke T, Kuhl J (1993) Representation of intentions: persisting activation in memory. J Exp Psychol Learning Memory Cogn 19:12111226 Graf P, Uttl B (2001) Prospective memory: a new focus for re- search. Conscious Cogn 10:437450 Haken H (1983) Synergetics. An introduction. Springer, Berlin Heidelberg New York Haken H (1987) Advanced synergetics. An introduction, 2nd edn. Springer, Berlin Heidelberg New York Haken H (1988/2000) Information and self-organization: a mac- roscopic approach to complex systems. Springer, Berlin Hei- delberg New York Haken H (1990) Synergetics of cognition. Springer, Berlin Hei- delberg New York Haken H (1991) Synergetic computers and cognition. Springer, Berlin Heidelberg New York Haken H (1996) Principles of brain functioning. Springer, Berlin Heidelberg New York Haken H, Kelso JAS, Bunz H (1985) A theoretical model of phase transitions in human hand movements. Biol Cybern 51:347356 Haken H, Portugali J (2003) The face of the city is its information. J Environ Psychol 23(4):382405 Kelso JAS (1995) Dynamic patterns. MIT Press, Cambridge Marsh RL, Hicks JL (1998) Event-based prospective memory and executive control of working memory. J Exp Psychol Learning Memory Cogn 24(2):336349 Marsh RL, Hicks JL, Hancock TW (2000) On the interaction of ongoing cognitive activity and the nature of an event-based intention. Appl Cogn Psychol 14:2941 Marsh RL, Hicks JL, Watson V (2002a) The dynamics of intention retrieval and coordination of action in event-based prospective memory. J Exp Psychol Learning Memory Cogn 28(4):652659 Marsh RL, Hancock TW, Hicks JL (2002b) The demands of an ongoing activity inuence the success of event-based prospec- tive memory. Psychon Bull Rev 9(3):604610 McDaniel MA, Einstein GO (2000) Strategic and automatic pro- cesses in prospective memory retrieval: a multiprocess frame- work. Appl Cogn Psychol 14:27144 Neisser U (1982) Memory observed: remembering in natural con- texts. WH Freeman, San Francisco Portugali J (ed) (1996) The construction of cognitive maps. Kluwer, Dordrecht Portugali J (1999) Self-organization and the city. Springer, Berlin Heidelberg New York Reese CM, Cherry KE (2002) The eects of age, ability, and memory monitoring on prospective memory task performance. Aging Neuro 9(2):98113 Sellen AJ, Louie G, Harris JE, Wilkins AJ (1997) What brings intentions to mind? A study of prospective memory. Memory 5(4):483507 Shannon CE, Weaver W (1949) The mathematical theory of communication. University of Illinois Press, Urbana Smith RE (2003) The cost of remembering to remember in event- based prospective memory: investigating the capacity demands of delayed intention performance. J Exp Psychol Learning Memory Cogn 29(3):347361 Smith G, Della Sala S, Logie RH, Maylor EA (2000) Prospective and retrospective memory in normal ageing and dementia: a questionnaire study. Memory 8(5):311321 Taylor RS, Marsh RL, Hicks JL, Hancock TW (2004) The inu- ence of partial-match cues on event-based prospective memory. Memory 12(2):203213 Van Der Berg SM, Aarts H, Midden C, Verplanken B (2004) The role of executive processes in prospective memory tasks. Eur J Cogn Psychol 16(4):511533 96 Vogels WWA, Dekker MR, Brouwer WH, de Jong, R (2002) Age- related changes in event-related prospective memory perfor- mance: a comparison of four prospective memory tasks. Brain Cogn 49(3):341362 West R, Herndon RW, Crewdson SJ (2001) Neural activity asso- ciated with the realization of a delayed intention. Cogn Brain 12(1):19 West R, Jakubek K, Wymbs N (2002) Age-related declines in prospective memory: behavioral and electrophysiological evi- dence. Neurosci Biobehav Rev 26(7):827833 97